---
type: "Evidence Item"
title: "Kimi K3 — HuggingFace Model Card (Moonshot AI)"
description: "Kimi K3’s official HuggingFace model card documents a 2.8T-parameter MoE model with 104B active parameters, scaling to 16-of-896 experts per token via a novel Stable LatentMoE."
resource: "https://huggingface.co/moonshotai/Kimi-K3"
tags: ["appendix-iii", "capabilities", "moonshot-ai"]
timestamp: "2026-07-27"
category: "capabilities"
publisher: "Moonshot AI"
cope_score: 83
confidence: 0.9
---

# Kimi K3 — HuggingFace Model Card (Moonshot AI)

# Claim

Kimi K3’s official HuggingFace model card documents a 2.8T-parameter MoE model with 104B active parameters, scaling to 16-of-896 experts per token via a novel Stable LatentMoE framework — yielding ~2.5× efficiency improvement over Kimi K2. Benchmark scores against frontier closed models: GPQA Diamond 93.5 (vs GPT-5.6 Sol 94.1, Claude Fable 5 92.6); DeepSWE 67.5 (vs GPT-5.6 Sol 73.0, Claude Opus 4.8 59.0); BrowseComp 91.2 (vs GPT-5.6 Sol 90.4, Claude Fable 5 88.0); MCPMark-Verified 94.5 (vs GPT-5.6 Sol 92.9, Claude Fable 5 87.4); SWE-Marathon 42.0 (SOTA; GPT-5.6 Sol 39.0, Claude Fable 5 35.0); Terminal-Bench 2.1 88.3 (vs GPT-5.6 Sol 88.8); OSWorld-Verified 84.8 (vs Claude Fable 5 85.0); JobBench 54.3 (vs Claude Fable 5 57.4, GPT-5.6 Sol 45.4); AutomationBench 30.8 (SOTA across all listed models). Architecture: 93 layers (1 dense + 92 MoE), 69 KDA + 24 Gated MLA attention layers, 7168-dim attention, 96 heads, vocabulary 160K tokens, native MXFP4 weight quantization with MXFP8 activations trained from SFT stage. Open weights released under Kimi K3 License.

# Relevance

Appendix III — capabilities: official technical specification and benchmark data for Kimi K3; cross-validates frontier-level autonomous performance (SWE-Marathon SOTA, AutomationBench SOTA, MCPMark-Verified SOTA) as open-weight model; JobBench score directly relevant to job displacement thesis

# Oracle Verdict

The HuggingFace model card provides the technical substrate confirming what the kimi.com blog post demonstrated qualitatively. The benchmark pattern is significant: Kimi K3 ties or exceeds frontier closed models (GPT-5.6 Sol, Claude Fable 5) across multiple agentic and autonomous task categories — SWE-Marathon (SOTA at 42.0), AutomationBench (SOTA at 30.8), MCPMark-Verified (SOTA at 94.5), BrowseComp (near-SOTA at 91.2) — and does so as an open-weight model activating only 104B of 2.8T parameters per token. The JobBench score (54.3) is the most directly displacement-relevant: it measures performance on realistic job tasks, and K3 outperforms GPT-5.6 Sol (45.4) while trailing only Claude Fable 5 (57.4) among listed models. The MXFP4 native quantization is architecturally notable — quantization-aware training from SFT stage means the compressed weights do not degrade agentic performance, making GPU-efficient deployment accessible at scale. Filed as [CAPABILITIES]: technical architecture and benchmark profile of the open-weight model substantiating frontier-level autonomous task performance.

# Metadata

* Publisher: Moonshot AI
* Category: capabilities
* Sector: AI research infrastructure / software engineering / agentic knowledge work / vision
* Capability: 2.8T-parameter MoE (104B active); Kimi Delta Attention (KDA) + Gated MLA architecture; 896 experts (16 active per token); 1M-token context; native multimodal (MoonViT-V2, 401M params); MXFP4/MXFP8 quantization-aware training; open weights under Kimi K3 License
* Cope score: 83
* Confidence: 0.9

# Related Concepts

* [Live evidence index](index.md)
* [Thesis](../thesis.md)

# Citations

[1] [Kimi K3 — HuggingFace Model Card (Moonshot AI)](https://huggingface.co/moonshotai/Kimi-K3)